Dissecting synaptic mechanisms of sound encoding in the mouse cochlea

Dissecting synaptic mechanisms of sound encoding in the mouse cochlea

Dissertation

for the award of the degree

“Doctor of Philosophy” Ph.D. Division of Mathematics and Natural Sciences

of the Georg August Universität Göttingen, within de doctoral program:

Sensory and Motor Neuroscience

of the Georg-August University School of Science (GAUSS)

submitted by Philippe Jean

From Nancy, France Göttingen, 2019

Members of the Thesis Advisory Committee:

Prof. Dr. Tobias Moser

Molecular anatomy, physiology, and pathology of sound coding and prosthetics, Institute for Auditory Neuroscience & InnerEarLab, University Medical Center Göttingen

Prof. Dr. Thomas Dresbach

Synaptogenesis group, Department of Anatomy and Embryology, Georg-August University, Göttingen

Dr. Andreas Neef

Biophysics of neural computation, Department of Nonlinear Dynamics and Network Dynamics group, Georg-August University, Göttingen

Further members of the Examination Board:

Prof. Dr. Martin Göpfert

Department of cellular neurobiology, Georg-August University, Göttingen

Dr. Manuela Schmidt

Somatosensory Signaling and Systems Biology Group, Max Planck Institute for Experimental Medicine, Göttingen

Dr. Katrin Willig

Optical Nanoscopy in Neuroscience, Max Planck Institute for Experimental Medicine, Göttingen

Date of oral examination 27/02/19

Declaration

Herewith I declare that this thesis has been written independently and with no other sources and aids than quoted.

Philippe Jean

Göttingen, January 7, 2019

List of Contents

List of abbreviations ...

1. Abstract ... 1

2. Introduction ... 3

2.1 Sound and hearing ... 3

2.2 The mammalian ear ... 3

2.2.1 The outer, middle and inner ear ... 3

2.2.2 The cochlea ... 3

2.2.3 The organ of Corti ... 6

2.2.4 Outer hair cells ... 6

2.2.5 Inner hair cells ... 7

2.2.5.1 Inner hair cell ribbon synapses ... 8

2.2.5.2 Voltage gated Ca²⁺ channels and their subunits ... 10

2.2.5.3 Ca²⁺ signaling at inner hair cell active zones ... 15

2.2.6 Encoding of sound intensity in spiral ganglion neurons ... 19

2.2.7 Spatial heterogeneity of synaptic Ca²⁺ influx properties at inner hair cells ... 21

2.3 Aim of this work ... 23

3. Chapter 1 “The synaptic ribbon is critical for sound encoding at high rates and with temporal precision” ... 25

4. Chapter 2 “Pou4f1 defines a subgroup of type I spiral ganglion neurons and is necessary for normal inner hair cell presynaptic Ca²⁺ signaling” ... 74

5. Chapter 3 “Intrinsic planar polarity mechanisms influence the position-dependent regulation of synapse properties in inner hair cells” ... 104

6. Discussion ... 137

6.1 Cellular mechanisms of wide dynamic range of sound encoding ... 138

6.1.1 Presynaptic candidate mechanisms ... 138

6.1.2 Postsynaptic candidate mechanisms ... 141

6.1.3 Efferent candidate mechanisms ... 144

6.2 Outlook ... 146

7. References ... 148

Acknowledgements ... 163

List of abbreviations

ABR auditory brainstem responses

ACh acetylcholine

AID α interacting domain

AMPA α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid AVCN anteroventral cochlear nucleus

AZ active zone

BAPTA 1,2- bis (2-aminophenoxy) ethane-N,N,N9,N9-tetraacetate

CaBP calcium binding protein

CaM calmodulin

CDI calcium-dependent activation

Cf characteristic frequency

C.M. center of mass

ΔCm membrane capacitance changes

CtBP2 C-terminal binding protein 2

DPOAE distortion product otoacoustic emissions

EGTA ethylene gly- col-bis-(2-aminoethyl)-N,N,N',N'-tetraacetic acid

EM electron microscopy

EPSC excitatory post-synaptic current EPSP excitatory post-synaptic potential F0 fluorescence at resting state

ΔF fluorescence change

ΔFmax maximal fluorescencechange

FS freeze substitution

FV fluorescence-voltage relationship

FWHM full width half maximum

GPCR G protein coupled receptor

HC hair cell

HPF high pressure freezing

IHC inner hair cell

IV current-voltage relationship

k slope

KO knockout

L.A. long axis

LJP liquid junction potential

MET mechanoelectrical transduction

MVR multivesicular release

MP membrane proximal

NPMC non parametric multiple comparison

OHC outer hair cell

PCP planar cell polarity

PD presynaptic density

PSD postsynaptic density

PTXa pertussis toxin catalytic subunit

QCa calcium charge

RA ribbon associated

RIM rab3-interacting molecule

RIM-BP RIM binding protein

RRP readily releasable pool

Rseries series resistance

S.A. short axis

S.D. standard deviation

S.E.M. standard error of the mean

SGN spiral ganglion neuron

SNARE soluble N-ethylmaleimide-sensitive factor attachment protein receptors

SR spontaneous rate

STED stimulated emission depletion

SV synaptic vesicle

TAMRA carboxytetramethylrhodamine

UVR univesicular release

VDI voltage-dependent inactivation VGCC voltage-gated Ca²⁺ channel Vglut vesicular glutamate transporter Vh voltage for half activation

1

1 Abstract

Cochlear inner hair cells (IHCs) are the genuine sensory receptors that translate sound-borne cochlear vibrations into neuronal signals via ribbon synapses with the spiral ganglion neurons (SGNs). The precise mechanisms of these ribbon synapses, the first relay of the auditory pathway, are still not fully resolved. Sound intensity coding over a wide dynamic range is thought to be fractionated through the SGNs presenting distinct firing characteristics. One hypothesis is that much of this diversity reflects the presynaptic heterogeneity observed among the active zones (AZs) of IHCs. There, a single AZ is the sole input to its associated SGN and it has been demonstrated that the AZs present very distinct properties according to their position in the IHCs.

The AZs facing the spiral limbus (modiolar side) show bigger ribbons correlated with a stronger Ca²⁺ influx activated at more depolarized membrane potentials than those at the side facing the pillar cells (pillar side). Differences in the voltage-dependent activation of the Ca²⁺ channels are an attractive explanation of the diverse postsynaptic spontaneous rates, where the SGNs associated with modiolar AZs with more depolarized activation range present a low spontaneous rate, while the pillar side targeting SGNs present a higher spontaneous rate. In this thesis, I first focused on deciphering the role of the synaptic ribbon at the first auditory synapse. Together with collaborators, we characterized the morphology and physiology of the ribbonless synapses by immunofluorescence and electron microscopy as well as patch–clamp/Ca²⁺ imaging of IHCs and systems physiology. We demonstrated a compensatory reorganization of the presynapses into several small ribbonless AZs, indicated a regulation of presynaptic Ca²⁺ influx by the ribbon and revealed a corresponding threshold increase as well as an impaired vesicle replenishment in the absence of the synaptic ribbon. The second part of my thesis aimed to decrypt the mechanisms setting the position-dependent heterogeneous properties of IHC AZs, putatively contributing to the wide dynamic range of sound encoding. Performing immunostainings and patch-clamp recordings combined with fast live Ca²⁺ imaging, we tested two different candidate mechanisms.

We firstly investigated if the transcription factor Pou4f1, expressed nearly entirely in a type I SGN subpopulation targeting the IHC modiolar face. We suggested that Pou4f1 defines a subset of low spontaneous rate, high threshold SGNs by decreasing the presynaptic voltage sensitivity leading to a depolarized shift of Ca²⁺ influx activation of their associated modiolar AZs. Then, we focused on the planar polarity mechanisms dictating hair bundle orientation and apical surface asymmetry, and proposed a role for the Gαi/LGN complex in regulating the position-dependent AZ properties in IHCs.

2

Introduction

3

2 Introduction

2.1 Sound and hearing

Sound is the principal mean of communication among animals but is also essential for their survival by indicating the location of prey or predators and by supporting navigation. Sound refers to a periodic, elastic compression and rarefaction of the transmitting medium (gas, liquid or solid). Sound hence propagates in the form of longitudinal waves (but spherically from the source) that carry two main properties, their amplitudes and frequencies. A third property, the localization can be detected thanks to a pair of ears, evaluating differences in the arrival time (low frequency sounds) and intensity (high frequency sounds) of the sound stimulus between these both ears, called interaural time and level differences. Depending on the distance between these two ears, the wave propagation angle and the constant interaural axis, time differences of a few µs need to be detected to decipher sound origin. This requires an incredible fidelity in the transmission of sound-evoked neural signaling in the nervous systems. Moreover, the ear can process acoustic stimuli that widely range in frequency and amplitude, and mammals can encode sound pressures ranging over six orders of magnitude. The fundamental mechanisms responsible for the precise, dynamic and indefatigable sound encoding are far from being completely understood.

2.2 The mammalian ear

2.2.1 The outer, middle and inner ear

The mammalian ear is divided into the outer, middle and inner ear. The outer ear is composed of a first part, the auricle focusing and filtering the incoming sound to the second part, the ear canal. At the end of this canal, the sound waves vibrate the tympanic membrane transferring the sound into the middle ear, a cavity filled with air. The sound wave is then carried by three ossicles (malleus, incus, and stapes) to the oval window of the cochlea in the inner ear, that also houses the vestibular labyrinth that senses head position and movement (Figure 1A).

2.2.2 The cochlea

The cochlea is a snail shaped structure consisting of a bony core around which several turns (which number depends on the species) of three fluid-filled compartments are coiled up (Figure

Introduction

4

1A). These compartments are the scalae vestibuli and tympani filled with perilymph (containing a comparably low amount of K⁺ (~5 mM), and the scala media, filled with endolymph that has a much higher concentration of K⁺ (~160 mM) due to secretion of K⁺ from the stria vascularis, lining the lateral wall and forming the outside boundary of the scala media (Figure 1B). When the stapes strikes the oval window, an amplification of the mechanical stimulus is made by both the larger surface of tympanic membrane as compared to the oval window membrane, and the leverage effects of the ossicles, which is essential for not reflecting the sound at the scala vestibuli fluid which has much greater impedance than air. Since this fluid is incompressible, the increase in pressure in the upper compartment, the scala vestibuli, is transmitted towards the lower compartment, the scala tympani resulting in a vibration of the basilar membrane and finally an outward bulging of the round window membrane. Both scalae are connected at the helicotrema at the apex of the cochlea. This system of pressure transfer causes a vertical travelling wave on the basilar membrane along the length of the cochlear turns (Figure 1C).

The basilar membrane is narrow and rigid at the base and wide and soft at the apex of the cochlea. These anatomical characteristics determine the inertia and stiffness of different parts of the cochlea, whereby defining their impedance. In consequence, high frequencies result in maximal movement of the basilar membrane in the basal cochlea, whereas low frequencies best vibrate the apical basilar membrane. This leads to a tonotopic frequency mapping along the length of the cochlea, covering the entire hearing range of the organism (Figure 1C).

Cochlear amplification for soft sounds and nonlinear compression for strong sounds are thought to enable the auditory system to grade the output of the cochlea for changes in input despite the limited dynamic range of sensory and neural mechanisms. Knocking out α-tectorin, detaching the tectorial membrane, generate a linear and monotonic compression of the basilar membrane.

Therefore, the tectorial membrane was proposed to ensure proper outer hair cell (OHC) response (see below) to basilar membrane motion via their anchorage, required for the compression (Legan et al., 2000). The receptor organ for hearing, called organ of Corti, is sandwiched between the tectorial and basilar membranes (Figure 1D).

Introduction

5

Figure 1: Outer, middle and inner ear

(A) Schematic representing the external ear that focuses and filters the incoming sound to the ear canal.

At the end of this canal, the waves conduct vibrations onto the tympanic membrane transferring the sound into the middle ear cavity. The sound wave is then carried by three ossicles to the oval window of the inner ear, consisting of two parts: the cochlea and the vestibular system. (B) Schematized cross section of a cochlea showing the arrangement of 3 coiled up fluid filled compartment that are the scala vestibuli, tympani and media. In this last compartment resides the organ of Corti surmounting the basilar membrane. (C) Simplified representation of an uncoiled cochlea and its basilar membrane. The sound causes a travelling wave of the basilar membrane. Each frequency of stimulation maximally excites a particular position of the basilar membrane. (D) Schematic of the organ of Corti, consisting of one row of IHCs, three rows of OHCs and several types of supporting cells, altogether sandwiched between the tectorial and basilar membranes. Modified from (Kandel, 2012).

2.2.3 The organ of Corti

The organ of Corti is a highly organized structure in the cochlea responsible for the processing of the sound. Situated on top of the basilar membrane, it consists of one row of IHCs, three

Introduction

6

rows of OHCs and several types of supporting cells such as Hensen’s cells, Deiter’s cells, phalangeal cells, inner and outer pillar cells. While the hair cells (HCs) are responsible for the sensory processing, the supporting cells play important roles in homeostasis and structural support of the organ of Corti. The HCs are sandwiched between the basilar and tectorial membranes (Figure 1D). The cuticular plates of the HCs possess groups of stereocilia (so-called hair bundles) made of a dense-filament core, arranged in two to three rows and are anchored in the tectorial membrane. The stereocilia are inter-connected at their apical tips by tip links (Pickles et al., 1984), filamentous structures mainly composed mainly of cadherin 23 (Siemens et al., 2004; Söllner et al., 2004) and protocadherin 15 (for review, see Müller, 2008). They are thought to be directly linked to the mechanoelectrical transduction (MET) channels (Assad et al., 1991; Howard and Hudspeth, 1988) situated at the top of the stereocilia (Beurg et al., 2009;

Jaramillo and Hudspeth, 1991; Lumpkin and Hudspeth, 1995). During sound-borne vibrations of the basilar membrane, the vertical displacement of the organ of Corti leads to a shearing movement, causing the deflection of the stereocilia, which in turn gates the MET channels. The tip links between stereocilia were suggested to work as springs ruled by Hook’s law. A displacement towards the longest stereocilia causes increased tension in the tip links, which in turn opens the MET channels. The increase or decrease of cation influx from the endolymph into the cells, mainly K⁺, through these channels depolarizes or hyperpolarizes the cell, respectively (Corey and Hudspeth, 1979).

2.2.4 Outer hair cells

The active membrane electromotility of the cylindrical OHCs (and potentially active bundle movements) mechanically amplify the vibration of the cochlear partition for soft sounds. The lateral membrane of the OHCs is densely packed with prestin, an integral membrane protein which belongs to the SLC26 family of ion transporters (Lohi et al., 2000; Oliver et al., 2001;

Zheng et al., 2000). This protein exhibits conformational changes upon voltage variations, resulting in stretching or shrinking of the cell, which are convened by interaction with chloride and bicarbonate anions in the cytoplasm (Oliver et al., 2001). Prestin-driven electromotility and, potentially, active bundle movements greatly amplify the vibration of the basilar membrane at a given tonotopic position for soft sounds with high frequency selectivity (Chan and Hudspeth, 2005b; Kennedy et al., 2005; Liberman et al., 2002; Zheng et al., 2000; and for review see Hudspeth, 2008).

Introduction

7

Figure 2: Stereocilia and tip links

The stereocilia are made of dense actin filaments and are inter-connected by tip links, mostly composed of cadherin 23 and protocadherin 15. They are thought to directly connect to the MET channels. During sound detection, the vertical displacement of the basilar membrane leads to the deflection of the stereocilia, which in turn gate the MET channels. A displacement towards the longest stereocilia causes increased tension in the tip links, which in turn opens the MET channels. The increase or decrease of cation influx from the endolymph into the cells through these channels depolarizes or hyperpolarizes the cell, respectively. Not to scale. Modified from (Lukacs, 2016).

2.2.5 Inner hair cells

Similarly to the OHCs, there are hair bundles arranged in two to three rows of stereocilia on the cuticular plate of the IHCs as well (for review, see Raphael and Altschuler, 2003). Each IHC possesses 20-50 stereocilia (depending on species and location along the basilar membrane), each presenting a diameter of ~250 nm and being inter-connected via their tip links. During sound-borne vibrations of the basilar membrane, while the hair bundles of the OHCs are deflected, the resulting movement of fluid radially displaces the hair bundles of the IHCs, permitting influx of cations. As for OHCs, this leads to a quick depolarization of the IHC membrane potential, and a shift towards the smallest stereocilia row causes a slackening of the

Introduction

8

tip links, resulting in closing of MET channels and a more negative membrane potential. The depolarization activates voltage-gated calcium (Ca²⁺) channels (VGCCs) at the ribbon synapses, ultimately leading to an entry of Ca²⁺ ions inside the cell, thereby triggering exocytosis of glutamate into the synaptic cleft. However, the precise synaptic transmission machinery is still enigmatic due to the peculiar features making the IHC ribbon synapse unique (Figure 3).

2.2.5.1 Inner hair cell ribbon synapses

The ribbon synapses formed by IHCs and postsynaptic SGNs are the first relay of the auditory pathway. IHCs must be able to detect sudden sound pressure changes in the environment and to convey the signal faithfully to the downstream neurons. To support the incessant stimulation of these receptors by the environment, the synapses have to maintain high rates of sustained release. Therefore, vesicle recycling has to be rapid, efficient and indefatigable. The hallmark feature to which ribbon synapses owe their name is a proteinaceous structure called the synaptic ribbon, which can be found at the synapses of auditory and vestibular hair cells, as well as photoreceptors, retinal bipolar cells, and pinealocytes (for review, see Fuchs et al., 2003; Lenzi and von Gersdorff, 2001; Sterling and Matthews, 2005).

At mouse IHC synapses, the ribbons are identified as spherical, ellipsoidal, or bar- shaped electron-dense body at the electron microscopy (EM) level. The size of these bodies varies among and within cells (spanning from 100 to 500 nm in the mouse IHCs). By molecules of unknown identity, a large number of synaptic vesicles (SVs, increasing with the size of the ribbon) is tethered to the ribbon structure forming a SV-halo around the ribbon body (Figure 3).

The main and core-component of the ribbon is RIBEYE (Schmitz et al., 2000), a protein unique to ribbon synapses. Large agglomerates of RIBEYE compose around two-third of the ribbon (for review, see Zanazzi and Matthews, 2009). RIBEYE has an N-terminal A domain and a C- terminal B domain. The A domain is thought to have a structural role, whereas the B domain is probably exposed in the cytosol, and is therefore suggested to have a metabolic function.

Interestingly, the B domain is identical to the C-terminal binding protein 2 (CtBP2), a transcription factor ubiquitously found in most tissues, making the specific deletion of the ribbon difficult. Despite decades of work, the role of the synaptic ribbon in sensory cells is still not completely resolved. The role of the ribbon has been widely investigated including approaches that employed natural modification of ribbon size or abundance (Hull et al., 2006;

Introduction

9

Mehta et al., 2013), photoablation (Mehta et al., 2013) and genetic manipulation removing or altering the ribbons (Frank et al., 2010; Khimich et al., 2005; Lv et al., 2016; Sheets et al., 2017). Lately, specific knock out of the Ribeye A domain achieved complete lack of ribbons from the mouse retina (Maxeiner et al., 2016) and organ of Corti (Becker et al., 2018; Jean et al., 2018). A long-standing hypothesis poses that the ribbon operates as a conveyor belt replenishing release sites by facilitated diffusion of SVs on the ribbon surface towards the site of consumption and promotes SV priming (Becker et al., 2018; Frank et al., 2010; Graydon et al., 2014; Jean et al., 2018; Mehta et al., 2013; Zenisek et al., 2000, for a deviating view see Jackman et al., 2009, where the ribbon is a safety belt slowing down the exocytosis in cones).

The ribbon has also been proposed to establish a large complement of vesicular release sites and Ca²⁺ channels at the AZ (Frank et al., 2010), to ensure close spatial coupling of Ca²⁺

channels and vesicular release sites (Maxeiner et al., 2016), or enhancing presynaptic Ca²⁺

signals by limiting diffusional Ca²⁺ spread (Graydon et al., 2014). Finally, the ribbon has also been proposed to contribute to multivesicular release (MVR) by coordinated or homotypic fusion potentially explaining the strong heterogeneity in size and shape of the spontaneous excitatory postsynaptic currents (EPSCs). However, this mechanism has been questioned recently, and univesicular release (UVR) has been proposed instead, where a single vesicle could either engage a full collapse or a pore flickering fusion (Chapochnikov et al., 2014;

Grabner and Moser, 2018; Huang and Moser, 2018).

Several other scaffold proteins have also been identified in ribbon synapses. CAST and ELKS were shown to be indispensable for synapse development, function and maintenance in the mouse retina (tom Dieck et al., 2012; Hagiwara et al., 2018). The selective deletion of long Piccolo isoforms left the retinal ribbon synapses unchanged (Regus-Leidig et al., 2013), while the knock-down of Piccolino, a short ribbon-specific Piccolo variant, disrupted the presynaptic ribbon plate-shaped morphology (Regus-Leidig et al., 2014). The ribbon anchoring role of Bassoon has been put in evidence by the reduced number of anchored ribbons and the presence of cytosolic floating ribbons observed in the mutant mouse (Dick et al., 2003; Frank et al., 2010;

Khimich et al., 2005). Moreover, super resolution stimulated emission depletion microscopy (STED) described the misalignment of Ca²⁺ channel clusters, even in the ribbon-anchored synapses of Bassoon mutant mice. Finally, membrane capacitance (Cm) measurements exposed a reduction of the fast component of exocytosis, reflecting the readily releasable pool (RRP).

All together, these synaptic alterations resulted in an impaired auditory function (Buran et al., 2010; Jing et al., 2013). Whether the reduction of the exocytosis is due to the loss of Bassoon or to the loss of ribbon still needs to be clarified. No major alteration of the exocytosis at IHCs

Introduction

10

has been described by Cm measurements in IHCs of Ribeye knockout (KO) mice where the synaptic ribbons were completely removed, likely due to compensation involving the formation of several small ribbonless AZs in IHCs (Becker et al., 2018; Jean et al., 2018). Nonetheless, analysis of sound encoding in vivo revealed altered SV release and refilling phenotype at high rates of stimulation. Interestingly, IHC synapses appear to operate independently of the classical neuronal soluble N-ethylmaleimide-sensitive factor attachment protein receptors (SNAREs) SNAP-25, syntaxin-1, and synaptobrevin-1 or synaptobrevin-2 (Nouvian et al., 2011).

Moreover, unlike glutamatergic neurons mainly using vesicular glutamate transporter isoforms Vglut1 or Vglut2, IHCs express Vglut3 (Ruel et al., 2008; Seal et al., 2008, also found in glial cells and non-glutamatergic neurons). Furthermore, synaptophysin, a synaptic vesicle marker, as well as the Ca²⁺ sensors synaptotagmins I and II were not detected in the mature IHCs (Reisinger et al., 2011; Safieddine and Wenthold, 1999), for controversy see (Beurg et al., 2010;

Johnson et al., 2010). The VGCCs have also been shown to be localized at the AZ below the ribbon (Brandt et al., 2005), however the detailed mechanisms by which all these protagonists interact with each other is still poorly resolved. Most importantly, the subunit composition of the VGCC clusters is of great importance to shape the Ca²⁺ signal at this atypical synapse.

2.2.5.2 Voltage gated Ca²⁺ channels and their subunits

Ca²⁺ signal plays many central roles in cells. At conventional synapses the neurotransmitter release is principally caused by Ca²⁺ influx through P/Q-, N-, and/or R-type VGCCs. Unlike those conventional synapses, the ribbon synapses in IHCs are not governed by action potentials, but driven by a graded potential and are therefore equipped with specific mechanisms regulating the VGCCs to faithfully transduce the signal into neurotransmission.

Remarkably, the predominant Ca²⁺ channel isoform found in IHCs is the L-type CaV1.3 (90%

of the Ca²⁺ channels, while the rest possibly comprise different kind of L-type (CaV1.4), as well as R-type (CaV2.3) channels (Brandt et al., 2003). Hence, IHCs exhibit dihydropyridine- sensitive Ca²⁺ currents (a class of drugs which can have either inhibitory (Isradipine, Nifedipine) or augmenting effects (BayK8644) on the channel opening probability) that have fast activation kinetics, open at comparably negative potentials (between −60 and −50 mV), and exhibit very little inactivation (Cui et al., 2007), rendering them suitable to support sustained exocytosis (Brandt et al., 2003; Dou et al., 2004; Platzer et al., 2000). CaV1.3 Ca²⁺

channels are clustered at the ribbon synapses (Brandt et al., 2005; Roberts et al., 1990; Zenisek

Introduction

11

et al., 2003), leaving a relatively low number of extrasynaptic Ca²⁺ channels (Brandt et al., 2005; Neef et al., 2018). Consequently, mice lacking the CaV1.3 Ca²⁺ channels are congenitally deaf (Platzer et al., 2000), due to failure to trigger exocytosis (Brandt et al., 2003).

Figure 3: Inner hair cell active zone

Illustration of the synaptic ribbon (magenta), selected cytomatrix AZ proteins, and CaV1.3 Ca²⁺ channels that group at the base of the ribbon. The scaffolding proteins Bassoon (the anchor of the ribbon) and RIM are involved in clustering the synaptic Ca²⁺ channels through direct or indirect interactions. RIM binding proteins (RIM-BPs) interact with both RIM and Bassoon and are molecular linker candidate between Ca²⁺ channels and vesicular release sites. RIM further interacts with SVs through Rab3 interaction. Not to scale.

Based on fluctuation analysis of Ca²⁺ currents, it has been proposed that mature mouse IHCs at the apex possess on average about 1,700 VGCCs (Brandt et al., 2005). Immature IHCs are thought to contain more Ca²⁺ channels (Beutner and Moser, 2001; Brandt et al., 2003;

Johnson et al., 2005; Wong et al., 2014; Zampini et al., 2010). Assuming an extra-synaptic density of 1 channel/µm², the average number of CaV1.3 channels per AZ was thus estimated to be ≈ 80 in mature apical IHCs (Brandt et al. 2005). In a later study, using two different Ca²⁺

imaging approaches via microiontophoresis and non-stationary fluctuation analysis, the number of Ca²⁺ channels per AZ was spanning from 30 to 330 (mean of 125) and 20 to 300 (mean of 78), respectively (Neef et al., 2018). The VGCCs consist of one main pore-forming

Introduction

12

α1 subunit, assembling with β, γ, α2 and δ auxiliary subunits that shape the trafficking and functional properties of the channel (Figure 4) (Ertel et al., 2000), and can be arranged into five different types depending on their current properties and drug sensitivity (for review, see Tsien et al., 1995). L-type Ca²⁺ channels exhibit long-lasting and large currents. They are high- voltage activated and can be blocked by dihydropyridines, phenylalkylamines and benzothiazepines. P/Q- and N- type channels also activate at high voltages but have a lower conductance than L-type. P/Q-type are sensitive to w-Aga IVA toxin (Llinás et al., 1989), while N-type channels are sensitive to w-conotoxin GVIA (Nowycky et al., 1985). R-type channels activate at high voltages as well (Niidome et al., 1992) but are resistant to toxins (Williams et al., 1994), and T-type channels are a class of low-voltage activated channels presenting only small and rapidly inactivating currents (Carbone and Lux, 1984; Nilius et al., 1985).

The CaVα1 subunits are large multi transmembrane domains transcribed from ten different genes. Each of their four homologous domains are linked to each other by cytoplasmic loops. The S4 segment of each domain contains multiple positively charged residues and is thought to act as voltage sensor (Glauner et al., 1999; Logothetis et al., 1992), whereas the S5 and S6 segments form the pore region. Furthermore, interaction sites with other Ca²⁺ channel subunits and molecules are present. Multiple C-terminal splice variants of CaV1.3 Ca²⁺ channels were identified, translated to long and short CaV1.3 isoforms, exhibiting different activation, inactivation and opening probability (Bock et al., 2011; Shen et al., 2006;

Singh et al., 2008; Tan et al., 2011).

Introduction

13

Figure 4: Subunit composition and structure of voltage-gated Ca²⁺ channels

Schematic illustrating the subunit composition and structure of VGCCs. The CaVα1 subunit is composed of four domains, each of which contains six transmembrane segments. The voltage sensor is thought to be located in segment S4. The CaVβ subunit is intracellular and binds to the CaVα1 subunit via the α- interaction domain. CaVα2δ is composed of two parts, the membrane-bound CaVδ and the extracellular CaVα2. CaV is a transmembrane subunit. Modified from (Catterall et al., 2005).

CaVβ subunits are entirely intracellular domains encoded by four different genes, co- expressed with the CaVα1 subunits and crucial for their trafficking and function (reviewed in Buraei and Yang, 2010). They bind to the intracellular loop between domains I and II of the CaVα1 subunit (Pragnell et al., 1994). These subunits are part of the membrane-associated guanylate kinase proteins family (Hanlon et al., 1999), which organize intracellular signaling pathways with multiple protein-protein interaction domains. Therefore, CaVβ subunits have been associated to various modulatory effects. They were proposed to increase the transport of CaVα1 to the plasma membrane via the α interacting domain (AID) (Bichet et al., 2000; Gebhart et al., 2010); to increase the open probability of the Ca²⁺ channel (Neely et al., 1993); and to change their voltage-dependence of activation (Jones et al., 1998) and inactivation (Meir and Dolphin, 2002). In IHCs, the β2 subunit is the main CaVβ subunit and its genetic deletion involved deficits in cochlear amplification and sound encoding. IHCs of mice lacking CaVβ2 in all tissues, but the heart, exhibited a reduced membrane expression of CaV1.3 Ca²⁺ channels, resulting in a decreased synaptic Ca²⁺ influx and exocytosis. Moreover, CaVβ2 are essential for development since the efferent innervation persisted onto the IHCs (Neef et al., 2009).

Introduction

14

Apparently, CaVβ3 and CaVβ4 are also expressed in IHCs (Kuhn et al., 2009; Neef et al., 2009) but only a mutation of CaVβ4 affected CaV1.3 expression at the plasma membrane (Kuhn et al., 2009).

The CaVα2δ subunits are members of the glycosylphosphatidylinositol-anchored family that are expressed throughout the nervous system. They are encoded from four genes and consist of two parts, CaVα2 and CaVδ, translated from the same mRNA and post-translationally cleaved.

The CaVδ subunit is a single transmembrane segment and linked to the extracellular CaVα2

subunit by a disulfide bond (Davies et al., 2010). They were shown to associate with other CaVα1 subunits such as CaVα1A, CaVα1B, and CaVα1C (Liu et al., 1996; Witcher et al., 1993). The main function of CaVα2δ is to augment the abundance of CaVα1 at the plasma membrane (reviewed in Dolphin, 2013). This increase in was shown to result from a reduced turnover of the channels resulting in a higher stability at the membrane and a chaperone-like function of CaVα2δ (Bernstein and Jones, 2007; Gao et al., 2000; Hoppa et al., 2012). A mutant mouse line lacking the expression of full length CaVα2δ2, exhibited mild hearing impairment and altered cochlear amplification along with decreased Ca²⁺ currents and voltage sensitivity in IHCs.

These results proposed that CaVα2δ2 is the predominant subunit of CaV1.3 in IHC (Fell et al., 2016).

CaVγ subunits are encoded from height different genes and are believed to possess four transmembrane domains (Green et al., 2001; Jay et al., 1990), however their role has not been as much elucidated as those of the other subunits. The most consistent effect of VGCC modulation by CaVγ is the induction of a positive shift of voltage-dependence of activation (reviewed in Black, 2003). However, their role is still controversial due to their rare detection as part of VGCCs. These subunits also regulate α-amino-3-hydroxy-5-methyl-4- isoxazolepropionic acid (AMPA) receptor trafficking, localization and modulation (Nicoll et al., 2006; Osten and Sternbach, 2006).

VGCCs show two types of inactivation: voltage-dependent inactivation (VDI) and Ca²⁺

dependent inactivation (CDI; Brehm and Eckert, 1978). VDI comes from the capability of VGCCs to inactivate in response to voltage-dependent conformational changes. The structural components of VGCCs responsible for VDI depend on interactions between the pore of the VGCC and the AID region of the cytosolic I-II loop of CaVα1 (Dafi et al., 2004; Stotz and Zamponi, 2001). The VDI response can be read as a response of the AID-CaVβ complex to block the channel pore (Tadross et al., 2010). The amount of VDI differs markedly depending

Introduction

15

on the CaVα1 and β subunits. Interestingly, CaV1.3 channels exhibit an additional shield region within the S6 segment and are therefore more resistant to VDI (Tadross et al., 2010).

Additionally, VGCC inactivate upon Ca²⁺ entry. This CDI restricts excessive Ca²⁺ influx in response to electrical activity. Calmodulin (CaM) preassociates with L-, P/Q-, and R-type Ca²⁺

channels even in the absence of Ca²⁺ (Erickson et al., 2001; Pitt et al., 2001) and, once Ca²⁺

bound, is responsible for CDI by binding to the IQ-motif of the CaVα1 subunit (Peterson et al., 1999; Qin et al., 1999; Zühlke et al., 1999). CaM possesses two distinct Ca²⁺ binding regions at the N-terminal that senses global increases in Ca²⁺ concentration (Cens et al., 2006), and C- terminal lobe senses local, sharp Ca²⁺ concentrations rising in the Ca²⁺ domain at the mouth of the channel (DeMaria et al., 2001; Yang et al., 2006). CDI can be regulated in a variety of ways.

It can be influenced by the diverse types and alternative splicing of CaVα1 subunits (Shen et al., 2006), by the type of CaV subunit co-expressed with the channel as well as by other EF-hand Ca²⁺ binding proteins (Cui et al., 2007).

2.2.5.3 Ca²⁺ signaling at inner hair cell active zones

The molecular mechanisms responsible for anchoring the Ca²⁺ channels at the IHC AZ have not yet been fully elucidated. IHCs lacking Bassoon exhibited an impaired Ca²⁺ channel clustering at the AZs, reducing the whole-cell and synaptic Ca²⁺ influx. However, a direct interaction was not found and the involvement of the loosened ribbon anchorage in this phenotype was not addressed (Frank et al., 2010). A promising candidate is rab3-interacting molecule (RIM). RIMs are multi-domain proteins that positively regulate the number of Ca²⁺ channels at the presynaptic AZ (Gebhart et al., 2010; Han et al., 2015; Kaeser et al., 2011; Kintscher et al., 2013; Kiyonaka et al., 2007). It has been shown that RIM2α and RIM2β promote a large complement of synaptic Ca²⁺ channels at IHC AZs and are required for normal hearing. Indeed, the AZs of RIM2α-deficient IHCs were shown to cluster fewer synaptic CaV1.3 Ca²⁺ channels, resulting in reduced synaptic Ca²⁺ influx associated with a reduction of exocytosis, while the apparent Ca²⁺ dependence of exocytosis was unchanged. Moreover, auditory brainstem responses (ABRs) indicated a mild hearing impairment (Jung et al., 2015). Afterwards, an interaction of the C2B domain of RIM2α and RIM3γ with the C-terminus of the pore-forming α–subunit of CaV1.3 channels was reported, positively regulating their plasma membrane expression in HEK293 cells (Picher et al., 2017a). The interacting partner of RIM, RIM binding protein 2 (RIM-BP2), was also shown to positively regulate the number of synaptic CaV1.3

Introduction

16

channels and thereby facilitate SV release. Its disruption resulted in a mild deficit of synaptic sound encoding (Krinner et al., 2017).

Interestingly, VGCCs can likely be opened at resting membrane potentials (Robertson and Paki, 2002), and additionally Ca²⁺ ions can continuously enter the cell through MET channels (Lumpkin and Hudspeth, 1995). To be able to retransmit with high fidelity the sound- evoked stimulation, the IHCs need to spatiotemporally modulate the Ca²⁺ signals with great precision. Ca²⁺ ATPase at the cell membrane, as well as thapsigargin sensitive intracellular stores and mitochondria supposedly contribute to the regulation of cytosolic [Ca²⁺] at least in immature IHCs (Kennedy, 2002). However, these mechanisms may not be sufficiently fast and localized close enough to the AZs to rapidly terminate sharp [Ca²⁺ ]i increase mediated by synaptic Ca²⁺ channel clusters (Roberts, 1994). Therefore, IHCs are equipped with mobile proteinaceous Ca²⁺ buffers, parvalbumin-α, calretinin and calbindin (Hackney et al., 2005).

Their concentrations as well as their Ca²⁺ association and dissociation kinetics play a major role in determining both the amplitude and the spatiotemporal properties of the presynaptic Ca²⁺

signals that govern neurotransmitter release (Edmonds et al., 2000; Neher and Augustine, 1992;

Roberts, 1993). They have been proposed to regulate presynaptic IHC function for metabolically efficient sound coding (Pangrsic et al., 2015). IHCs lacking all three proteins showed excessive exocytosis during prolonged depolarizations, despite enhanced Ca²⁺

dependent inactivation of their Ca²⁺ currents. Synaptic sound encoding was largely unaltered, suggesting that the excess exocytosis primarily occurred extrasynaptically (Pangrsic et al., 2015).

Ca²⁺ binding proteins (CaBPs), too, are members of the EF-hand Ca²⁺ binding proteins family. They are encoded by eight genes and subdivided into two subfamilies with distinct intracellular localization profiles (Haeseleer et al., 2000). All of them are comprised of two lobes, each bearing a pair of EF-hands connected by an inter-lobe flexible a-helix linker region.

Unlike CaM, EF-hand in the CaBP N-lobe is not binding Ca²⁺. In contrast to ubiquitous CaM expression in all eukaryotic cells, CaBPs are restricted to sensory cells and neurons (Haeseleer et al., 2000). CaBPs might perturb the normal binding of CaM to VGCCs, probably by competing with binding to the IQ domain and preventing CDI (Cui et al., 2007; Schrauwen et al., 2012; Yang et al., 2006). Given the advantage of CaBPs over CaM that results in preventing CDI despite similar Kd-values, CaBPs can be assumed to reside in closer proximity to the IQ- domain. Whether the competition between CaBPs and CaM for the IQ domain depends only on Ca²⁺ requires further investigation (Tadross et al., 2010). Among CaBPs, CaBP1 is the most

Introduction

17

heavily investigated isoform in respect to VGCC function: CaBP1 upregulates Ca²⁺ dependent facilitation and voltage-dependent inactivation of CaV1.2 channels (Oz et al., 2011; Zhou et al., 2004). To date only CaBP4 was reported to be required for normal vision, was localized to photoreceptor synaptic terminals and, apparently, is involved in rod and cone synapse formation and function (Haeseleer et al., 2004; Zeitz et al., 2006). In addition, CaBP5 is expressed in rod bipolar cells and some cone bipolar cell types, but is not strictly required for vision (Rieke et al., 2008). IHCs display considerably weak CDI in comparison with neurons or cell lines. In mouse IHCs, CaBP1-5 were detected by immunohistochemistry, but their respective function is under current investigation. CaBP2 was proven to be a deafness gene (Schrauwen et al., 2012;

Picher et al., 2017b). Cabp2 loss showed impaired ABRs, enhanced Ca²⁺ channel inactivation and reduced spontaneous and sound-evoked firing rates. CaBP2 was proposed to inhibit CaV1.3 Ca²⁺ channel inactivation, and thus sustains the availability of CaV1.3 Ca²⁺ channels for synaptic sound encoding (Picher et al., 2017b).

Together, the clustering of VGCCs at AZs and the Ca²⁺ binding to mobile Ca²⁺ buffers establish domains of locally elevated [Ca²⁺ ]i. For any effector of [Ca²⁺]i, like the vesicular Ca²⁺ sensor for fusion, their spatial relation to these domains is of considerable functional significance. In mature IHCs, Cav1.3 Ca²⁺ channels localize at mature presynaptic AZs underneath synaptic ribbons. This tight organization importantly decreases the distance that incoming Ca²⁺ has to travel to reach the Ca²⁺ sensors on the synaptic vesicles and thus allows for high sensitivity to changes in membrane potential. In the scenario of nanodomain control of exocytosis (Neher, 1998), in which the opening of one or few Ca²⁺ channels would trigger the release of a vesicle, the Ca²⁺ sensor is located very close to the channel(s) (≈ 20 nm), and will be exposed to sharp [Ca²⁺ ]i increment of several tens to hundreds of µM upon channel opening (Roberts, 1994). Experimentally, due to the small coupling distance, it can be characterized by the fact that only very ‘fast’ buffers (high association rate, kon; e.g. 1,2- bis (2-aminophenoxy) ethane-N,N,N9,N9-tetraacetate (BAPTA); (Naraghi, 1997) are effective in capturing Ca²⁺ ions before they reach the Ca²⁺ sensor and hence in inhibiting stimulus- secretion coupling. One advantage is that the nanodomain control would allow to act rapidly upon channel opening, reflected in exocytic delays in the range of hundreds of microseconds (Yamada and Zucker, 1992). Exposing the Ca²⁺ sensor of fusion to the saturating [Ca²⁺] of the nanodomain, would make the timing of synaptic sound encoding less sensitive to stimulus intensity, but would be less reliable since the opening of Ca²⁺ channels is stochastic (Moser et al., J Physiol 2006). In the microdomain control of exocytosis (Augustine et al., 2003; Neher, 1998), the opening of several Ca²⁺ channels trigger the release of one vesicle, the Ca²⁺ sensor

Introduction

18

is placed farther away from the channels (≈ 100-200 nm), and will experience smaller [Ca²⁺ ]i

elevations, a longer time to peak [Ca²⁺ ]i, and an increased buffering susceptibility due to the longer coupling distance. It can be characterized by the effectiveness of ‘slow’ buffers (lower association rate, kon; e.g. ethylene gly- col-bis-(2-aminoethyl)-N,N,N',N'-tetraacetic acid (EGTA) (Naraghi, 1997). One advantage is that this microdomain control is more reliable than the nanodomain control since it builds on more channels, but is slower since it needs more Ca²⁺ influx to reach the required local [Ca²⁺] at the sensor. However, if several channels in a cluster open more or less simultaneously, the Ca²⁺ increments would sum up, and the required concentration at the sensor would be reached more quickly. Thus, rapid exocytosis is also achievable with a microdomain control of transmitter release, as it is the case at immature the calyx of Held synapse (Borst and Sakmann, 1996).

Those two extreme scenarios (Figure 5) differ in term of apparent Ca²⁺ cooperativity of exocytosis when the the number of open Ca²⁺ channels is varied. A linear increase in the exocytic response is expected for a nanodomain control of exocytosis due to the dominance of one Ca²⁺ channel in governing the fusogenic Ca²⁺, whereas a supra-linear increase is predicted for a microdomain control reflecting intrinsic high cooperativity of exocytosis (Matveev et al., 2009). The preferred scenario of coupling between Ca²⁺ channels and sensors at IHC synapses is the nanodomain-like control, due to evidences found by correlating presynaptic Ca²⁺ influx with exocytosis (Brandt et al., 2005; Pangrsic et al., 2015; Wong et al., 2014) or EPSC charges (Goutman and Glowatzki, 2007; Keen and Hudspeth, 2006). These post synaptic responses are carried out by the SGNs that convey the sound information via the auditory nerve to the brain stem.

Introduction

19

Figure 5: Coupling between voltage-gated Ca²⁺ channels and vesicular release sites (A) Nanodomain control of exocytosis: The Ca²⁺ sensors are located very close from the Ca²⁺ channels (~ 20 nm), and therefore would be exposed to sharp [Ca²⁺]i increment upon channel opening. The opening of one or few channels would trigger the release of a vesicle. In such a scenario, the vesicle release is linearly dependent on the number of open Ca²⁺ channels. (B) Microdomain control of exocytosis: The Ca²⁺ sensors are located further away from the Ca²⁺ channels (~100-200 nm), and will be exposed to smaller [Ca²⁺ ]i elevations. The opening of more Ca²⁺ channels would trigger the release of a vesicle. Hence, the release of vesicles is non-linearly dependent on the number of open Ca²⁺

channels. Not to scale.

2.2.6 Encoding of sound intensity in spiral ganglion neurons

In brief, mature IHCs are connected with about 10 to 20 type I SGNs (hereafter “SGN”) in a 1:1 connection. These SGNs are the first afferent neurons of the auditory system, and make up around 95% of the ganglion, while the remaining ones are type II and contact individually a dozen of OHCs (Kiang et al., 1982; Spoendlin, 1969, 1972). The IHC graded receptor potential governs the Ca²⁺ dependent release of neurotransmitter that activates glutamate receptors on the afferent bouton resulting in excitatory postsynaptic potentials (EPSPs) in the SGN (Corey and Hudspeth, 1979; Robertson and Paki, 2002; Sewell, 1984). When a large enough number of receptors have been activated, the resulting currents depolarize the cell beyond threshold,

Introduction

20

thus triggering an action potential. The function of these synapses is similar to an analog- digital converter, in which an analog sensory signal is encoded as a digital code. The first heminode was reported to reside just beneath habenulae perforatae not far from the IHCs suggesting that the action potential is triggered nearby the small postsynaptic bouton, reducing the duration of membrane charging to the threshold of spiking. These features guarantee that SGNs can respond to the presynaptic stimulus with great temporal accuracy. The other important feature of SGNs is their phasic excitability. SGNs only spike once or few times due to their expression of voltage-gated K⁺-(Kv)channels. This together with refractoriness makes them able to filter out some other EPSPs which are not elicited by the well-timed presynaptic releasing of neurotransmitters, thereby enhancing the precision of signal transmission in SGNs (for review see Davis and Crozier, 2016; Rutherford and Moser, 2016).

During spontaneous firing (spontaneous rate: SR, measured as the firing rate in the absence of sound), more than 90% of the spontaneous EPSPs can generate an action potential (Rutherford et al., 2012). The spiking rate of SGNs increases up to several hundreds of Hz in response to increasing sound intensity and is essentially limited by the neuron refractoriness.

However, mammalian auditory system can encode sound pressures varying by 6 orders of magnitude. Despite the nonlinear compression of basilar membrane, SGNs still need to be able to respond to sound pressures that vary over thousand-fold. In mammalian cochlea, SGNs vary in their response to sound stimulation. Their firing rate, response threshold and dynamic range are highly heterogeneous, even among neurons showing similar Cf (indicating that they are innervating the same IHC or ones in close proximity) (Liberman, 1978). Therefore, SGNs work as different channels to the brain enabling the response of wide ranges of sound pressure.

Three response patterns for sound levels of the SGNs were identified: “saturating,” “sloping- saturation,” and “straight” modes. The SGNs for the saturating mode have the lowest acoustic threshold (~10–30 dB) to fire action potentials but their firing rate does not increase after a certain sound intensity. The fibers presenting sloping-saturation (getting slower) and straight mode (linear) have higher thresholds; however, their firing rate never saturate. The SR showed negative relationship with response threshold in cat (Liberman, 1978), guinea pig (Winter et al., 1990), gerbil (Ohlemiller and Echteler, 1990), rat (Barbary, 1991) and mouse (Taberner and Liberman, 2005). It was found that, in cat, SGNs can be roughly divided into three groups based on their SR (high, medium, low), each corresponding to a different sensitivity (Liberman, 1978). Later, it was observed that the medium- and low-SR rate (higher threshold) neurons preferentially innervate the neural/modiolar (facing afferent fibers) side of IHCs, while high SR neurons innervated the abneural/pillar (facing OHCs) side (Liberman, 1982).

Introduction

21

These findings suggest the existence of a segregation of neural responses properties at the synaptic level.

Figure 6: Innervation pattern of mature cochlear hair cells

Illustration displaying the innervation pattern of cochlear hair cells in mature mice. Each IHC is the sole input of a dozen of type I SGNs. Fewer type II SGNs individually transmit signal from several OHCs.

Efferent axons largely and directly innervate the OHCs, whereas this innervation is sparser at the level of the IHCs, and make input with the afferent. Modified from (Kandel, 2012).

2.2.7 Spatial heterogeneity of synaptic Ca²⁺ influx properties at inner hair cells

One attractive mechanism driving the synapse-based segregation of SGN firing behaviors is the diversity of IHC AZ properties. First studies (Frank et al., 2009; Meyer et al., 2009) established confocal Ca²⁺ imaging at IHCs of mice after the onset of hearing using low-affinity Ca²⁺

indicator and strong buffering condition allowing to characterize spatially confined Ca²⁺

domains. These localized domains of high [Ca²⁺ ] also occur at IHC synapses studied at

Introduction

22

physiological-like Ca²⁺ concentrations and buffering (Neef et al., 2018). Several pieces of evidence strongly suggested that these observed Ca²⁺ microdomains arose from Ca²⁺ influx through synaptic CaV1.3-Ca²⁺ channel clusters at the ribbon synapses. First, they occurred at sites marked by a ribbon-labeling CtBP2/RIBEYE-binding fluorescent peptide (Zenisek et al., 2004). Moreover, These Ca²⁺ microdomains were not observable when removing extracellular Ca²⁺. Additionally, their voltage-dependence mimicked that of the whole-cell Ca²⁺ current.

Finally, inhibition of Ca²⁺ induced Ca²⁺ release by ryanodine or intracellular Cs⁺ did not reduce the amplitude of the Ca²⁺ microdomains. A great variation in the voltage dependence (voltage of half-maximal activation) of the Ca²⁺ fluorescence increment among AZs of an individual IHC was shown, which was four times larger than that of the whole-cell current compared among the same experiments. In addition to these shifts in the activation, a large fluorescence variability of Ca²⁺ microdomains has also been found for strong depolarizations when Ca²⁺

channel activation saturates. Finally, the study did not support a major contribution of Ca²⁺

signal augmentation by local differences in Ca²⁺ buffering/sequestration, or the Ca²⁺ channel open probability among IHC synapses (Frank et al., 2009). Each presynaptic AZ is controlled by a common IHC receptor potential and provides the sole excitatory input to “its” postsynaptic SGN. ²⁺ This study proposed that the IHC differently adjusts the gating and number of Ca²⁺

channels at each AZ, establishing distinct voltage dependences of transmitter release, being a putative presynaptic mechanism for driving different SGN firing characteristics.

To spatially characterize these observations, a later study (Ohn et al., 2016) used a spinning disk confocal microscope, a multi-beam scanning system permitting fast 3D-live- imaging allowing an analysis of most if not all AZ properties as a function of position within an individual IHC. In parallel, a semi quantitative confocal microscopy of immunolabelled IHC synapses was performed. The study suggested that IHCs spatially decompose sound intensity information into different outputs by varying the maximal Ca²⁺ influx and the voltage- dependent activation among their AZs along the pillar/modiolar IHC axis. These two influx characteristics exhibited opposite spatial gradients. Modiolar AZs showed bigger ribbons, more Ca²⁺ channels, and on average, a tendency towards a stronger maximal Ca²⁺ influx than the AZs from the opposite side. A greater number of Ca²⁺ channels is expected to trigger more spontaneous and evoked release, and a stronger spontaneous and evoked firing rate at the postsynapse (Robertson and Paki, 2002). However, they found that modiolar AZs, on average, presented a voltage operating range shifted towards more positive potentials as compared to the pillar side AZs, consistent with the idea that modiolar synapses drive low SR- high-threshold SGNs (Figure 7). These AZs holding more Ca²⁺ channels may support higher maximal rates of

Introduction

23

transmitter release, which could explain the higher susceptibility of their associated afferent fibers to excitotoxic insult during acoustic overexposure.

However, whereas innervation of functionally distinct classes of type I SGNs at the level of IHC was established in cats, data revealing the presynaptic heterogeneity were obtained from mice where a different afferent connectivity is conceivable. Moreover, an immaturity phenotype cannot be ruled out for several of the mouse studies since the experiments were done shortly after the hearing onset. Indeed a recent study based on immunostaining analysis showed an inversion of the ribbon and AMPAR patch modiolar/pillar gradients from P21 of age, gradient being maximal and unchanged from P28 (Liberman and Liberman, 2016). Additionally, the variability in the operating voltage range of Ca²⁺ influx among synapses of one-half of the cell (modiolar or pillar) could exceed the difference between the average pillar and modiolar synapses. The SGNs with identical frequency tuning but different sound-response properties apparently receive input from the same IHC and collectively vehicle acoustic information across the entire audible range of sound pressures to the brain. Most probably a combination of presynaptic, postsynaptic and efferent mechanisms underlie this diversity of SGN sound- response properties.

2.3 Aim of this work

The first part of my thesis focused on assessing the role of the synaptic ribbon in sound encoding. Taking advantage of mice KO for Ribeye A domain and consequently lacking ribbons, I together with collaborators characterized the morphology of the ribbonless synapses by immunofluorescence and electron microscopy. The physiology of sound encoding was then studied by performing patch–clamp/ Ca²⁺ imaging of IHCs and in vivo extracellular recordings of SGNs. The second part of my thesis aimed to decipher the cellular mechanisms contributing to the diversity of SGN firing behaviors thought to underlie the wide dynamic range of sound encoding. Performing immunostainings and patch-clamp recordings combined with fast live Ca²⁺ imaging, and with help from collaborators, I tested two different candidate mechanisms. I investigated if: 1) Pou4f1, a type I SGN transcription factor, postnatally expressed almost exclusively (~95%) in a subpopulation of SGNs targeting the modiolar face of the IHCs, and 2) the mechanisms involved in IHC planar polarity governing hair bundle orientation and apical surface asymmetry, are instructing the basolateral gradient of synapse properties, putatively responsible for the diverse firing properties of type I SGNs.

Introduction

24

Figure 7: Basolateral gradient of inner hair cell synapse properties and associated fiber firing

Illustration showing the IHC spatial synaptic connectivity. The low SR SGNs preferentially make a synapse with the modiolar side of the IHC, exhibiting large ribbons and Ca²⁺ channel clusters activated at more depolarized potentials than the ones on the pillar side, exhibiting smaller AZs and making contact with the high SR SGNs. Data taken from (Taberner and Liberman, 2005). Not to scale.

Chapter 1

25

3 Chapter 1: Published article

“The synaptic ribbon is critical for sound encoding at high rates and with temporal precision”

P. Jean^*, D. Lopez de la Morena^*, S. Michanski^*, L.M. Jaime Tobón^*, R. Chakrabarti, M.M.

Picher, J. Neef, S.Y. Jung, M. Gültas, S. Maxeiner, A. Neef,, C. Wichmann, N. Strenzke, C.

Grabner, and T. Moser

* Equal contributions

Author Contributions:

T.M., C.G., C.W., A.N., N.S. designed the study. P.J. performed immunohistochemistry and confocal immunofluorescence microscopy, patch-clamp, Ca²⁺ imaging, and computational modeling together with A.N.. D.L. performed in vivo extracellular recordings from single SGNs and AVCN neurons. S.M.

performed conventional embedding and EM of random and serial sections and contributed to high- pressure freezing. L.M.J.T. performed patch-clamp capacitance measurements. R.C. performed high- pressure freezing, freeze-substitution and electron tomography. M.M.P. performed patch-clamp capacitance measurements. J.N. performed immunohistochemistry, confocal and STED immunofluorescence microscopy, and co-supervised P.J. and L.M.J.T.. S.Y.J. performed immunohistochemistry and confocal immunofluorescence microscopy. M.G. contributed a statistical analysis of EM data. S.M. contributed mutant mice and genetic expertise, A.N. performed and supervised computational modeling. C.W. supervised EM and tomography. N.S. supervised in vivo extracellular recordings from single SGNs. T.M. co-supervised P.J. D.L. L.M.J.T. All authors performed analysis. T.M. and P.J. prepared the manuscript with contributions of all authors.

Chapter 1

Chapter 1

Chapter 1

Chapter 1

Chapter 1

Chapter 1

Chapter 1

Chapter 1

Chapter 1

Chapter 1

Chapter 1

Chapter 1

Chapter 1

Chapter 1

Chapter 1

Chapter 1

Chapter 1

Chapter 1

Chapter 1

Chapter 1

Chapter 1

Chapter 1

Chapter 1